Description
shared/python-dataviz.md (imported by charts-with-trending.md, in turn imported by ~15 daily-report workflows including api-consumption-report.md and github-mcp-structural-analysis.md) provisions its chart-rendering environment via python3 -m venv against the runner's system CPython — but that CPython requires GLIBC_2.38 while the agent sandbox ships 2.35, so the venv never loads. Two independent reports today hit this and fell back to hand-built static SVGs / skipped charts entirely: the GitHub API Consumption Report (#54501, "every CPython 3.10-3.14 install requires GLIBC_2.38... none load") and the MCP Structural Analysis (#54520, "the only usable interpreter is PyPy 3.9, and pandas/matplotlib/seaborn have no PyPy wheels"). A working fix already exists in this repo for the same root cause in a different workflow: daily-agentrx-trace-optimizer.md installs "a portable uv-managed CPython... under /tmp/gh-aw" specifically because "the agent sandbox cannot run the runner's CPython (glibc mismatch)".
Expected Impact
Restores real Python charting (matplotlib/pandas/seaborn/scipy) for every one of the ~15 workflows that import shared/python-dataviz.md, instead of each one silently falling back to hand-rolled SVGs, skipped charts, or ASCII approximations on a case-by-case basis.
Suggested Agent
Copilot coding agent — generalize the portable uv-managed CPython pattern already proven in daily-agentrx-trace-optimizer.md into shared/python-dataviz.md's setup step, replacing the python3 -m venv call.
Estimated Effort
Medium (1-4 hours)
Data Source
DeepReport analysis, 2026-08-21 cycle, cross-referencing discussions #54501 and #54520 (both hit the identical GLIBC mismatch independently) with the existing working fix in .github/workflows/daily-agentrx-trace-optimizer.md. No existing open issue found covering this specific shared-import gap.
Generated by 🔬 Deep Report · agent · 186.6 AIC · ⌖ 14.7 AIC · ⊞ 11.9K · ◷
Description
shared/python-dataviz.md(imported bycharts-with-trending.md, in turn imported by ~15 daily-report workflows includingapi-consumption-report.mdandgithub-mcp-structural-analysis.md) provisions its chart-rendering environment viapython3 -m venvagainst the runner's system CPython — but that CPython requiresGLIBC_2.38while the agent sandbox ships2.35, so the venv never loads. Two independent reports today hit this and fell back to hand-built static SVGs / skipped charts entirely: the GitHub API Consumption Report (#54501, "every CPython 3.10-3.14 install requires GLIBC_2.38... none load") and the MCP Structural Analysis (#54520, "the only usable interpreter is PyPy 3.9, and pandas/matplotlib/seaborn have no PyPy wheels"). A working fix already exists in this repo for the same root cause in a different workflow:daily-agentrx-trace-optimizer.mdinstalls "a portable uv-managed CPython... under/tmp/gh-aw" specifically because "the agent sandbox cannot run the runner's CPython (glibc mismatch)".Expected Impact
Restores real Python charting (matplotlib/pandas/seaborn/scipy) for every one of the ~15 workflows that import
shared/python-dataviz.md, instead of each one silently falling back to hand-rolled SVGs, skipped charts, or ASCII approximations on a case-by-case basis.Suggested Agent
Copilot coding agent — generalize the portable uv-managed CPython pattern already proven in
daily-agentrx-trace-optimizer.mdintoshared/python-dataviz.md's setup step, replacing thepython3 -m venvcall.Estimated Effort
Medium (1-4 hours)
Data Source
DeepReport analysis, 2026-08-21 cycle, cross-referencing discussions #54501 and #54520 (both hit the identical GLIBC mismatch independently) with the existing working fix in
.github/workflows/daily-agentrx-trace-optimizer.md. No existing open issue found covering this specific shared-import gap.